August Meetup: Interactive Audio

On the 4th of August we will host the next iteration of our monthly Audio Developer Meetup, where Robyn Pacetti from Protagonists! will be sharing their journey of creating a new kind of interactive audio experience. For our short talk, Fernando Garcia will present their latest research on detecting AI-generated content in hybrid music mixtures. As always, there will be a chance to network over drinks and pizza. Anyone interested in Audio development is welcome! Doors open at 18:00, talks start at 18:15. Please note: Attendance is strictly limited to people who have RSVP'd. Long talk details A journey to create interactive, imaginary audio worlds - Robyn Pacetti What if we could explore imaginary worlds without looking at a screen? Explore environments, engage with characters, be the protagonist — interact with an imaginary world you don't just listen to, but move through and shape through your choices. The founder of Protagonists! talks about her journey creating a new kind of interactive audio experience: the vision, the goals, how it works, what they've built so far, some of the challenges, and what doors this new kind of experience might open — for people and for the audio space. About the speaker Robyn Pacetti is the Founder of Protagonists!, an early-stage bootstrapped venture actively building a new kind of interactive audio entertainment: imaginary worlds you step into and shape through your choices — ScreenFree, no graphics. It's a return to the oldest way humans have told stories — by voice, with the listener imagining the world into being — but with agency, so you don't just listen, you act. After 25 years in the games industry (Amazon, Ubisoft, Wooga), Robyn now works at the frontier of interactive entertainment. Short talk details How Much AI Is in This Track? Detecting AI-Generated Content in Hybrid Music Mixtures - Fernando Garcia AI-generated music is reshaping the music industry. By early 2026, it made up around 44% of daily uploads on Deezer, and studies suggest most listeners can't tell it apart from human-made recordings. This has driven a major research effort on automatic detection, almost all of which asks one binary question: is this track AI or human? State-of-the-art detectors now hit over 99% accuracy on that question. But that isn't how AI actually shows up in modern production. Producers use it as a modular tool: an AI drum loop combined with a synthetic bassline and a human singer's voice. The result is hybrid material that's neither fully human nor fully AI, exactly the case the binary framing can't describe. This talk covers Fernando's work reframing detection for that reality. Instead of a yes/no label, they treat it as a regression problem: estimating a continuous AI energy ratio, the fraction of a track's acoustic energy that comes from AI stems. Fernando will explain why these detectors work at all (they exploit subtle spectral fingerprints left by neural audio codecs), and show how a model trained to estimate that ratio reveals patterns a binary detector misses entirely. The broader takeaway: near-perfect binary accuracy is not evidence that a detector is ready for realistic, mixed-content music. As AI becomes a normal part of production, the question worth asking isn't whether a track is AI, but how much. About the speaker Fernando Garcia de la Cruz is ML Engineer at seeqnc, working on AI detection. His interests range across generative models, music information retrieval, and real-time audio software, with a focus on building practical tools for musicians and technologists. He holds an MSc in Sound and Music Computing from the Music Technology Group (MTG), Universitat Pompeu Fabra in Barcelona, and a BSc in Electronics Engineering from Instituto Politécnico Nacional in Mexico City. He began his career as a software and data engineer in consultancy before moving into audio and music research, working on real-time audio tools and libraries at Dolby Laboratories, AI-generated music detection at BMAT, audio plugins at the MTG with the Essentia library, and most recently as a research intern at Sony CSL Paris. He has also spoken at the Audio Developer Conference and contributes to the Sound of AI community.